Are you using to for speech-to-text/text-to-speech? I have been using SpeechRecognizer & SpeechSynthesizer and they have been pretty underwhelming.